A knowledge management system based on ITSM system knowledge graph technology

Abstract

The present invention provides a knowledge management system based on ITSM system knowledge graph technology, including at least one set of underlying architecture and at least one set of knowledge architecture, both of which are oriented to heterogeneous graph architecture; the underlying architecture includes a knowledge generation unit, a database, and a comprehensive unit, and the knowledge generation unit, the database, and the comprehensive unit are used as nodes to form a first heterogeneous graph, and the knowledge architecture includes a program module set in the database and operated by the comprehensive unit to realize visual multidimensional knowledge management and learning, which is used to construct a second heterogeneous graph, thereby forming a first heterogeneous network of the first heterogeneous graph and a second heterogeneous network of the second heterogeneous graph in all geographical regions, and finally establishing an attention mechanism in the first heterogeneous network and the second heterogeneous network respectively, completing the operation and maintenance of the underlying architecture, and the knowledge management of the knowledge architecture. The operation and maintenance, and intelligent management of the underlying architecture, the knowledge architecture, and the hybrid architecture are realized.

Description

Technical Field

[0001] The present invention relates to a heterogeneous graph knowledge management system, and in particular to a knowledge management system based on ITSM system knowledge graph technology, and belongs to the field of intelligent management. Background Art

[0002] The existing IT Service Management (ITSM) system concept is general, without a unified technical architecture and positioning, but in general, it realizes its own service value through flexible updates, automation, and the use of artificial intelligence, becoming one of the means to improve the quality of information services. Since the application of knowledge is divided into five major directions: knowledge generation, knowledge storage, knowledge call, knowledge dissemination, and knowledge intelligent recommendation. Therefore, based on the concept of ITSM, we have restructured the above five major directions in knowledge management to obtain a heterogeneous graph knowledge management system, which is based on the ITSM system and will continue to deepen based on this solution in the future.

[0003] The benefit of heterogeneous graphs is that the objects of interest are formed into nodes, and different data objects can be constructed in the nodes, such as the data itself, data storage units, maintenance units, and analysis units. In addition, nodes can be associated with each other, so that dynamic updates between nodes can be achieved through the attention mechanism, thereby predicting user behavior and focus.

[0004] Therefore, facing the five major directions of knowledge as the underlying logic, the heterogeneous graph can be structured from the underlying logic, thereby realizing a new ITSM system. Summary of the invention

[0005] In order to solve the current problem of unclear concept of ITSM overall architecture, the present invention will provide a knowledge management system based on ITSM system knowledge graph technology, including at least one underlying architecture and at least one knowledge architecture both facing heterogeneous graph architecture, distributed in divided geographical areas;

[0006] The underlying architecture includes knowledge generation units that communicate with each other and can be used for users to input knowledge data, a database responsible for knowledge storage and maintenance, and a comprehensive unit responsible for knowledge query, call, dissemination, and intelligence. The knowledge generation unit, database, and comprehensive unit are used as nodes to form a first heterogeneous graph.

[0007] The knowledge architecture includes a program module set in the database and run by the integrated unit to realize visualized multi-dimensional knowledge management and learning, which is used to construct a second heterogeneous map, thereby forming a first heterogeneous network of the first heterogeneous map and a second heterogeneous network of the second heterogeneous map in all geographical areas, and finally establishing an attention mechanism in the first heterogeneous network and the second heterogeneous network respectively to complete the operation and maintenance of the underlying architecture and the knowledge management of the knowledge architecture, corresponding to the knowledge generation unit, the database, and the integrated unit. The knowledge management includes the management of knowledge generation, knowledge storage and maintenance, as well as the call, dissemination, and intelligence of knowledge, wherein the call and dissemination are executed according to the user's request, and the intelligence also includes the intelligence of the operation and maintenance, including regular updating of the first heterogeneous map, the second heterogeneous map, and the first heterogeneous network and the second heterogeneous network.

[0008] The first heterogeneous map includes the representation of each region formed by the geographical area division. in, is the area code, used to indicate the corresponding area. is the first node set, representing the knowledge generation unit, database, and integration unit, represents the edge between the first nodes, and the second heterogeneous graph includes a representation in each region formed by the geographical region division (The font is regular, note that it is different from the italic representation of the first node set) is the second node set, representing knowledge data, represents the edge between the second nodes, the first heterogeneous network is , the second heterogeneous network is , where the first heterogeneous network and the second heterogeneous network use the first heterogeneous graph and the second heterogeneous graph as the graph node set respectively. and As edges between graph nodes.

[0009] It is easy to understand that the graph is a subgraph of a heterogeneous network, which belongs to a heterogeneous network subgraph within each divided area and can perform intelligent operation and maintenance of the internal underlying architecture.

[0010] Furthermore, an attention mechanism is established in the first heterogeneous network and the second heterogeneous network, respectively, which specifically includes the following three steps:

[0011] The first step is to build the first attention mechanism in the first heterogeneous network to obtain the first normalization coefficients at the node, semantic, and subgraph levels. '

[0012] It is easy to understand that the more frequent the access, the more it reflects the superiority of the system in the area, which can be the popularity of the stored knowledge, the excellent system stability, the fast network speed, etc.

[0013] The second step is to construct a second attention mechanism in the second heterogeneous network to obtain the second normalization coefficients at the node, semantic, and subgraph levels. '

[0014] The third step is to construct a hybrid attention mechanism in the first heterogeneous network and the second heterogeneous network to obtain hybrid normalization coefficients at the node, semantic, and subgraph levels. The superscript Indicates blending.

[0015] Assume that the theoretical load used to characterize the system in each area is Current load Frequency of visits The comprehensive performance of the comprehensive indicators under the comprehensive indicators (i.e., the number of customers visiting and / or network traffic in a specified period of time), i.e., the performance function (1), where is the evaluation coefficient, when the label Take different numbers to represent The corresponding evaluation coefficient is is the coefficient, forming the system performance matrix in different areas , then the first attention mechanism and the second attention mechanism are respectively the first performance in considering the comprehensive ability performance (2) Second performance (3) For the hybrid representation of the hybrid attention mechanism, let the node, path, and subgraph be the third representation (4), where for The geographical region code where the node of interest is located, is the geographical area code where the beginning and end of the path are located, and the + sign is the geographical area code where other nodes on the path are located. is the geographical region code where the two nodes in the subgraph are located, and the + sign is the geographical region code where all nodes in the subgraph are located, 1, , is the number of the geographic area code, is the maximum number, i.e. the total number of geographical areas, the number is Medium Alternative .

[0016] It should be understood that the current load described in the present invention does not refer to the real-time load, but the load in a short period of time, such as one hour, one day, or one week or one month. The hybrid attention mechanism describes the relationship between the knowledge architecture and the underlying architecture, and at the same time indicates in which geographical areas the knowledge data is concentrated, and in which integrated units the service preferences of the knowledge data are generated. To a certain extent, it reflects the ITSM quality of the integrated unit, with users as the market to participate in the evaluation. The second and third items of the left matrix in the third expression are the weighted averages of the corresponding evaluation coefficients of each node. Replacing k with the number of the geographic area code does not change the value of the evaluation coefficient, but only switches the identity in different expressions.

[0017] The operation and maintenance method includes:

[0018] S1 monitors the performance matrix in real time based on the time the system is used by users.

[0019] S2 defines the change threshold of the performance matrix elements and the change threshold of the first to third performances. When real-time monitoring finds that the performance matrix has elements exceeding the corresponding change threshold, an alarm is issued, and / or when the first to third performances exceed the change threshold, the exceeding part is analyzed; wherein,

[0020] Analysis of excess includes first performance increment (5) Second performance increment (6) The third performance increment The summed items obtained by multiplying the matrices in (7) are monitored, and the values ​​of the items are compared with the change thresholds. If exceeded, the contribution distribution of the excess portion is calculated based on whether there are elements in the performance matrix that exceed the corresponding change threshold.

[0021] That is, the excess distribution is contributed by the elements in the performance matrix that exceed the corresponding change threshold (i.e., the elements of the left matrix in equations (5)-(7)) and the normalized coefficient increment, forming the contribution distribution of the two. Thus, real-time operation and maintenance can be performed at the two data levels of the underlying architecture and the knowledge architecture, thereby observing the stability of the system and the changes and development of knowledge.

[0022] The knowledge management includes recording the time and address of the user's generation, storage and maintenance, query, call, and dissemination of knowledge data to form a record table. Intelligent management includes re-correcting the attention mechanism based on the updated first heterogeneous graph, the second heterogeneous graph, and the first heterogeneous network and the second heterogeneous network, and analyzing the knowledge data that the user is interested in based on the user's records of generation, storage and maintenance, query, call, and dissemination of knowledge data for the user's reference.

[0023] The intelligent operation and maintenance includes re-correcting the attention mechanism according to the updated first heterogeneous graph, the second heterogeneous graph, the first heterogeneous network and the second heterogeneous network, and updating the evaluation coefficient using formula (1) according to the re-corrected attention mechanism, and analyzing the change rate of the first value third performance increment before and after the update for different users according to formulas (5)-(7), thereby adjusting the change threshold according to the change rate.

[0024] Optionally, the updating of the evaluation coefficient includes updating Any one of theoretical load, current load, access frequency, or a combination thereof.

[0025] It should be understood that the intelligent setting of this operation and maintenance is because if different users generally deviate from the change threshold, it means that the development of knowledge corresponding to the knowledge data is incomparable, and then we should consider updating the change threshold, which is specifically reflected in The changes in theoretical load, current load, and access frequency are statistically significant.

[0026] Preferably, the change threshold comprises a first change threshold characterizing a change in a representation matrix element, and a second change threshold characterizing a change in a first to third representation.

[0027] Preferably, the periodic updating of the first heterogeneous map, the second heterogeneous map, and the first heterogeneous network and the second heterogeneous network includes, within a predetermined time period and before the period, if the number of alarms exceeds a preset value, executing the update in advance. Beneficial Effects

[0028] The system is divided into the underlying architecture and the knowledge architecture, and heterogeneous graphs and heterogeneous graph networks are established separately to establish their respective attention mechanisms. The concept of representation is introduced to realize the operation and maintenance, as well as intelligent management of the underlying architecture, knowledge architecture, and hybrid architecture. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A knowledge management system based on ITSM system knowledge graph technology and its functional diagram according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] like Figure 1 , this embodiment provides a knowledge management system based on ITSM system knowledge graph technology, including at least one underlying architecture and at least one knowledge architecture both oriented to heterogeneous graph architecture, distributed in divided geographical areas, with ellipsis indicating multiple sets.

[0031] The underlying architecture (shown in a black chamfered box) includes knowledge generation units that communicate with each other and can be used for users to input knowledge data, a database responsible for knowledge storage and maintenance, and a comprehensive unit responsible for knowledge query, call, dissemination, and intelligence. The knowledge generation unit, database, and comprehensive unit are used as nodes to form a first heterogeneous graph.

[0032] The knowledge architecture (shown in the lake blue rounded box) includes a program module set in the database and run by the integration unit to realize visualized multi-dimensional knowledge management and learning, which is used to construct a second heterogeneous map, thereby forming a first heterogeneous network of the first heterogeneous map and a second heterogeneous network of the second heterogeneous map in all geographical regions, and finally establishing an attention mechanism in the first heterogeneous network and the second heterogeneous network respectively to complete the operation and maintenance of the underlying architecture and the knowledge management of the knowledge architecture, corresponding to the knowledge generation unit, the database, and the integration unit. The knowledge management includes the management of knowledge generation, knowledge storage and maintenance, as well as the call, dissemination, and intelligence of knowledge, wherein the call and dissemination are executed according to the user's request, and the intelligence also includes the intelligence of the operation and maintenance, including regular updates (indicated by the purple line) of the first heterogeneous map, the second heterogeneous map, and the first heterogeneous network and the second heterogeneous network.

[0033] The first heterogeneous map includes a representation in each region formed by the geographical region division ,in, is the area code, used to indicate the corresponding area. is the first node set, representing the knowledge generation unit, database, and integration unit, represents the edge between the first nodes, and the second heterogeneous graph includes a representation in each region formed by the geographical region division , (The font is regular, note that it is different from the italic representation of the first node set) is the second node set, representing knowledge data, represents the edge between the second nodes, the first heterogeneous network is represented as , the second heterogeneous network is expressed as , where the first heterogeneous network and the second heterogeneous network use the first heterogeneous graph and the second heterogeneous graph as the graph node set respectively. and As edges between graph nodes.

[0034] For example Figure 1 As shown, an attention mechanism is established in the first heterogeneous network and the second heterogeneous network, respectively, which specifically includes the following three steps:

[0035] The first step is to build the first attention mechanism in the first heterogeneous network to obtain the first normalization coefficients at the node, semantic, and subgraph levels. '

[0036] The second step is to construct a second attention mechanism in the second heterogeneous network to obtain the second normalization coefficients at the node, semantic, and subgraph levels. '

[0037] The third step is to construct a hybrid attention mechanism in the first heterogeneous network and the second heterogeneous network to obtain hybrid normalization coefficients at the node, semantic, and subgraph levels. The superscript Indicates blending.

[0038] Assume that the theoretical load used to characterize the system in each area is Current load Frequency of visits The comprehensive performance of the comprehensive indicators under the comprehensive indicators (i.e., the number of customers visiting and / or network traffic in a specified period of time), i.e., the performance function (1), where is the evaluation coefficient, when the label Take different numbers to represent The corresponding evaluation coefficient is is the coefficient, forming the system performance matrix in different areas , then the first attention mechanism and the second attention mechanism are respectively the first performance in considering the comprehensive ability performance (2) Second performance (3) For the hybrid representation of the hybrid attention mechanism, let the node, path, and subgraph be the third representation (4).

[0039] The operation and maintenance method includes:

[0040] S1 monitors the performance matrix in real time based on the time the system is used by users.

[0041] S2 defines a first change threshold for the change of the performance matrix elements and a second change threshold for the change of the performance matrix from the first to the third performance. When real-time monitoring finds that the performance matrix has elements exceeding the corresponding first change threshold, an alarm is triggered, and when the first to the third performance exceeds the change threshold, the exceeding part is analyzed; wherein,

[0042] Analysis of excess includes first performance increment (5) Second performance increment (6) The third performance increment The summed items obtained by multiplying the matrices in (7) are monitored, and the values ​​of the items are compared with the second change threshold. If exceeded, the contribution distribution of the excess portion is calculated based on whether there are elements in the performance matrix that exceed the corresponding second change threshold.

[0043] The system is capable of executing the above-mentioned knowledge management, specifically including recording the time and address of the user's generation of knowledge data, knowledge storage and maintenance, query, call, and dissemination, forming a record table, and managing intelligence including re-correcting the attention mechanism based on the updated first heterogeneous graph, the second heterogeneous graph, and the first heterogeneous network and the second heterogeneous network, and analyzing the knowledge data of interest to the user based on the user's generation of knowledge data, knowledge storage and maintenance, query, call, and dissemination records for the user's reference.

[0044] The intelligent operation and maintenance includes re-correcting the attention mechanism according to the updated first heterogeneous graph, the second heterogeneous graph, the first heterogeneous network and the second heterogeneous network, and updating the evaluation coefficient using formula (1) according to the re-corrected attention mechanism, and analyzing the change rate of the first value third performance increment before and after the update for different users according to formulas (5)-(7), thereby adjusting the change threshold according to the change rate.

[0045] The periodic updating of the first heterogeneous map, the second heterogeneous map, and the first heterogeneous network and the second heterogeneous network includes, within a predetermined time period and before a period, if the number of alarms exceeds a preset value, executing the update in advance.

Claims

1. A knowledge management system based on ITSM system knowledge graph technology, characterized in that: including at least one underlying architecture and at least one knowledge architecture, both oriented to a heterogeneous graph architecture, distributed into divided geographical regions; The underlying architecture includes knowledge generation units that communicate with each other and can be used for users to input knowledge data, a database responsible for knowledge storage and maintenance, and a comprehensive unit responsible for knowledge query, call, dissemination, and intelligence. The knowledge generation unit, database, and comprehensive unit are used as nodes to form a first heterogeneous graph. The knowledge architecture includes a program module set in the database and run by the integrated unit to realize visualized multi-dimensional knowledge management and learning, which is used to construct a second heterogeneous map, thereby forming a first heterogeneous network of the first heterogeneous map and a second heterogeneous network of the second heterogeneous map in all geographical areas, and finally establishing an attention mechanism in the first heterogeneous network and the second heterogeneous network respectively to complete the operation and maintenance of the underlying architecture and the knowledge management of the knowledge architecture, corresponding to the knowledge generation unit, the database, and the integrated unit. The knowledge management includes the management of knowledge generation, knowledge storage and maintenance, and the calling, dissemination, and intelligence of knowledge, wherein the calling and dissemination are performed according to the user's request, and the intelligence also includes the intelligence of the operation and maintenance, including regular updating of the first heterogeneous map, the second heterogeneous map, and the first heterogeneous network and the second heterogeneous network; The attention mechanism is established in the first heterogeneous network and the second heterogeneous network respectively, which specifically includes the following three steps: The first step is to build the first attention mechanism in the first heterogeneous network to obtain the first normalization coefficients at the node, semantic, and subgraph levels. , , Representation of; The second step is to build a second attention mechanism in the second heterogeneous network to obtain the second normalization coefficients at the node, semantic, and subgraph levels. , , Representation of; The third step is to construct a hybrid attention mechanism in the first heterogeneous network and the second heterogeneous network to obtain hybrid normalization coefficients at the node, semantic, and subgraph levels. , , The superscript Indicates blending. Assume that the theoretical load used to characterize the system in each area is , Current Load , visit frequency The comprehensive ability performance under the comprehensive index, that is, the performance function (1), where is the evaluation coefficient, when the label Take different numbers to represent , , The corresponding evaluation coefficient is , is the coefficient, forming the system performance matrix in different areas , then the first attention mechanism and the second attention mechanism are respectively the first performance in considering the comprehensive ability performance (2) Second performance (3) For the hybrid representation of the hybrid attention mechanism, let the node, path, and subgraph be the third representation (4), where for The geographical region code where the node of interest is located, is the geographical area code where the beginning and end of the path are located, and the + sign is the geographical area code where other nodes on the path are located. is the geographical area code where the two nodes in the subgraph are located, and the + sign is the geographical area code where all the nodes in the subgraph are located. is the number of the geographic area code, is the maximum number, i.e. the total number of geographical areas, the number is Medium Alternative .

2. The system according to claim 1, characterized in that in, The first heterogeneous map includes a representation in each region formed by the geographical area division ,in, is the area code, used to indicate the corresponding area. is the first node set, representing the knowledge generation unit, database, and integration unit, represents the edge between the first nodes, and the second heterogeneous graph includes a representation in each region formed by the geographical region division , is the second node set, representing knowledge data, represents the edge between the second nodes, the first heterogeneous network is , the second heterogeneous network is , where the first heterogeneous network and the second heterogeneous network use the first heterogeneous graph and the second heterogeneous graph as the graph node set respectively. and As edges between graph nodes.

3. The system according to claim 1, characterized in that The operation and maintenance method includes: S1 monitors the performance matrix in real time based on the time the system is used by users. , , ; S2 defines the change threshold of the performance matrix elements and the change threshold of the first to third performances. When real-time monitoring finds that the performance matrix has elements exceeding the corresponding change threshold, an alarm is issued, and / or when the first to third performances exceed the change threshold, the exceeding part is analyzed; wherein, Analysis of excess includes first performance increment (5) Second performance increment (6) The third performance increment The summed items obtained by multiplying the matrices in (7) are monitored, and the values ​​of the items are compared with the change thresholds. If exceeded, the contribution distribution of the excess portion is calculated based on whether there are elements in the performance matrix that exceed the corresponding change threshold.

4. The system according to any one of claims 1 to 3, characterized in that: The knowledge management includes recording the time and address of the user's generation, storage and maintenance, query, call, and dissemination of knowledge data to form a record table. Intelligent management includes re-correcting the attention mechanism based on the updated first heterogeneous graph, the second heterogeneous graph, and the first heterogeneous network and the second heterogeneous network, and analyzing the knowledge data that the user is interested in based on the user's records of generation, storage and maintenance, query, call, and dissemination of knowledge data for the user's reference.

5. The system according to claim 3, characterized in that The intelligent operation and maintenance includes re-correcting the attention mechanism according to the updated first heterogeneous graph, the second heterogeneous graph, the first heterogeneous network and the second heterogeneous network, and updating the evaluation coefficient using formula (1) according to the re-corrected attention mechanism, and analyzing the change rate of the first to third performance increments before and after the update for different users according to formulas (5)-(7), thereby adjusting the change threshold according to the change rate.

6. The system according to claim 5, characterized in that The updating of the evaluation coefficient includes updating , , theoretical load, current load, access frequency, or any one or a combination thereof.

7. The system according to claim 6, characterized in that The change thresholds include a first change threshold that characterizes a change in a representation matrix element, and a second change threshold that characterizes a change in a first to third representation.

8. The system according to claim 7, characterized in that The periodic updating of the first heterogeneous map, the second heterogeneous map, and the first heterogeneous network and the second heterogeneous network includes, within a predetermined time period and before a period, if the number of alarms exceeds a preset value, executing the update in advance.

Images